Marta Navarrete and Alfonso Araque
Instituto Cajal, Consejo Superior de Investigaciones Científicas. Madrid 28002, Spain.
Correspondence should be addressed to: Dr. Alfonso Araque, Instituto Cajal, Doctor Arce 37,
Madrid 28002. Spain.
Tel: 34-91-585 4710
Fax: 34-91-585 4754
E-mail: araque@cajal.csic.es
Short Title: Glia-mediated synaptic modulation by cannabinoids.
Keywords: Astrocytes, endocannabinoids, intracellular Ca
2+
, gliotransmitter release, glutamate, neuron-glia communication.
Acknowledgements:
We thank G. Perea, W. Buño, E.D. Martin, and L. Maglio for helpful comments, and A. Zimmer for the generous gift of the CB1R knock out mice. Supported by grants from Ministerio de Ciencia e Innovación (BFU2007-064764), Spain, European Union (HEALTH-
F2-2007-202167) and Cajal Blue Brain.

2
Endocannabinoids and their receptor CB1 play key roles in brain function. Astrocytes express
CB1Rs that are activated by endocannabinoids released by neurons. However, the consequences of the endocannabinoid-mediated neuron-astrocyte signaling on synaptic transmission are unknown. We show that endocannabinoids released by hippocampal pyramidal neurons increase the probability of transmitter release at CA3-CA1 synapses. This synaptic potentiation is due to CB1R-induced Ca 2+ elevations in astrocytes, which stimulate the release of glutamate that activates presynaptic metabotropic glutamate receptors. While endocannabinoids induce synaptic depression in the stimulated neuron by direct activation of presynaptic CB1Rs, they indirectly lead to synaptic potentiation in relatively more distant neurons by activation of CB1Rs in astrocytes. Hence, astrocyte calcium signal evoked by endogenous stimuli (neuron-released endocannabinoids) modulates synaptic transmission.
Therefore, astrocytes respond to endocannabinoids that then potentiate synaptic transmission, indicating that astrocytes are actively involved in brain physiology.

3
The endocannabinoid system is an important intercellular signaling system involved in a wide variety of physiological processes. Endocannabinoids and their receptor CB1 play relevant neuromodulatory roles in brain physiology (Chevaleyre et al., 2006; Freund et al., 2003) and are responsible for most of the psychotropic and behavioral effects of cannabinoids (Di Marzo et al.,
1994; Maldonado et al., 2006). They mediate retrograde inhibition of transmitter release, control neuronal excitability, and regulate short- and long-term synaptic plasticity (Chevaleyre et al., 2006;
Freund et al., 2003; Alger, 2002; Kreitzer and Regehr, 2002; Wilson and Nicoll, 2002). The effects of endocannabinoids and CB1R activation on synaptic transmission have been largely characterized in many brain regions (Chevaleyre et al., 2006; Heifets and Castillo, 2009; Kano et al., 2009;
Lovinger, 2008). Although a recent study in goldfish has shown that CB1R activation stimulates the release of dopamine that lead to the modulation of chemical and electrical synapses in the Mauthner cell (Cachope et al., 2007), endocannabinoid signaling is generally accepted to induce the inhibition of neurotransmitter release from presynaptic terminals (Chevaleyre et al., 2006; Hashimotodani et al., 2007).
Accumulating evidence has demonstrated the existence of bidirectional communication between astrocytes and neurons (Araque et al., 2001; Haydon and Carmignoto 2006; Nedergaard et al.,
2003; Perea et al., 2009; Volterra and Bezzi, 2002). Astrocytes respond with intracellular Ca
2+ elevations to neurotransmitters released by synaptic terminals (Perea and Araque, 2005; Porter and
McCarthy, 1996) and, in turn, modulate neuronal excitability and synaptic transmission by releasing neuroactive substances called gliotransmitters (Volterra and Bezzi, 2002; Araque et al., 1999;
Beattie et al., 2002; Perea and Araque, 2007).
Cannabinoid effects in brain physiology have been largely thought to be mediated by CB1Rs exclusively present in neurons. However, hippocampal astrocytes have recently shown to express functional CB1Rs that upon activation by endocannabinoids released from pyramidal neurons increase the astrocyte Ca
2+
levels and stimulate the release of glutamate, which, in turn, leads to the

4 modulation of the neuronal excitability through activation of NMDARs in neurons (Navarrete and
Araque, 2008). However, the consequences of the endocannabinoid-mediated neuron-astrocyte communication on synaptic transmission remain unknown.
Therefore, we used electrophysiological and Ca
2+
imaging techniques in mice brain slices to investigate the effects of the endocannabinoid-mediated neuron-astrocyte signaling on synaptic transmission at hippocampal synapses. We performed paired recordings from CA1 pyramidal neurons, monitored astrocyte Ca
2+
levels, and stimulated Schaffer collaterals using the minimal stimulation technique that activates single or very few presynaptic fibers (Dobrunz and Stevens,
1997; Isaac et al., 1996; Perea and Araque, 2007; Raastad, 1995). We found that endocannabinoids released by hippocampal pyramidal neurons increase the probability of transmitter release at CA3-
CA1 synapses by triggering a cascade of signaling events consisting in activation of CB1Rs in adjacent astrocytes, elevation of their Ca
2+
levels, stimulation of glutamate release and activation of presynaptic metabotropic glutamate receptor type 1. While endocannabinoids directly induced G i/o protein-mediated synaptic depression in synapses close to the source of endocannabinoids (i.e., synapses in the neuron that release endocannabinoids and nearby synapses) by activation of presynaptic CB1Rs, they indirectly led to synaptic potentiation in relatively more distant synapses
(i.e., synapses in other neurons not releasing endocannabinoids) by activation of astrocytes.
RESULTS
We first quantified the synaptic transmission properties of the excitatory postsynaptic currents
(EPSCs) in our experimental conditions (Figures 1A-1D). Synaptic responses showed failures and successes in neurotransmitter release (probability of release, Pr, was 0.31 ± 0.02; range, 0.07 - 0.83; n = 53 synapses), regular amplitude of successful responses (termed synaptic potency; 11.9 ± 0.6 pA; range, 4.1 - 25.5 pA; n = 53) and relatively low synaptic efficacy (i.e., mean amplitude of all

5 responses including failures: 7.7 ± 0.90 pA; range, 0.9 – 18.4 pA; n = 53) (cf. Dobrunz and Stevens,
1997; Isaac et al., 1996; Perea and Araque, 2007).
Then, we depolarized one pyramidal neuron to 0 mV for 5 s to stimulate endocannabinoid release (Chevaleyre and Castillo, 2004; Kreitzer and Regehr, 2001; Ohno-Shosaku et al., 2001;
Wilson and Nicoll, 2001), and monitored synaptic transmission at synapses in an adjacent neuron.
In 14 out of 51experiments (27 %), neuronal depolarization (ND) transiently increased Pr (from
0.24 ± 0.04 to 0.47 ± 0.05; n = 14; P < 0.01) without changing the synaptic potency (from 12.1 ±
1.2 pA to 13.0 ± 2.0 pA; n = 14; P = 0.57), which resulted in a transient enhancement of the synaptic efficacy (from 4.8 ± 1.6 pA to 9.2 ± 1.6 pA; n = 14; P < 0.01) (Figure 1E). These effects were associated with changes in the paired-pulse facilitation, suggesting a presynaptic origin, and were reliably evoked by successive ND (Figure S1).
In other 14 out of the 51 cases, ND evoked a transient decrease of Pr (from 0.30 ± 0.06 to 0.18 ±
0.04; n = 14; P < 0.05) without changing the synaptic potency (from 10.8 ± 3.8 pA to 9.7 ± 2.7 pA; n = 14; P = 0.35), which led to a transient decrease of the synaptic efficacy (from 5.3 ± 1.2 pA to
2.8 ± 0.7 pA; n = 14; P < 0.01) (Figure 2). In synapses that showed ND-evoked depression of synaptic transmission in control, subsequent perfusion with the CB1R antagonist AM251 (2µM) abolished this effect, which is similar to the depolarization-induced suppression of excitation
(DSE), and which is due to activation of presynaptic CB1Rs (Chevaleyre et al., 2006; Kreitzer and
Regehr, 2001; Ohno-Shosaku et al., 2002).
We then investigated if endocannabinoids mediated the synaptic potentiation (e-SP) of transmitter release. Because the appearance of ND-evoked DSE and e-SP were reliably induced, to characterize pharmacologically both phenomena we first identified specific synapses undergoing either DSE or e-SP in control conditions and then analyzed the effects of subsequent superfusion of pharmacological agents on those identified synapses (unless stated otherwise). The ND-evoked e-
SP observed in control recordings was abolished after perfusion with the CB1R antagonist AM251

6
(n = 6; Figures 1C, 1D and 1F), strongly suggesting that the ND-evoked e-SP is mediated by endocannabinoids released from the postsynaptic neuron.
Furthermore, e-SP was absent when the depolarized neuron was dialyzed with 40 mM BAPTA
(Chevaleyre et al., 2006; Kreitzer and Regehr, 2001; Navarrete and Araque, 2008; Ohno-Shosaku et al., 2002; Wilson and Nicoll, 2001) (which prevents endocannabinoid release; n =11; Figures 1F and 1G), or in slices from transgenic mice lacking CB1Rs (Zimmer et al., 1999) (n = 13; Figures 1F and 1G). Likewise, DSE was also absent in these conditions (n = 11 and 13 synapses in BAPTAfilled neurons and transgenic mice, respectively) (Figures 2E and 2F). Although no identifying control recordings were possible in these experiments, the total absence of ND-evoked e-SP or DSE further supports the idea that they are mediated by endocannabinoids released from the postsynaptic neuron.
We then investigated whether action potential (AP) generation could similarly evoke e-SP and
DSE. We recorded the stimulating neuron in current-clamp conditions and applied depolarizing pulses that generated trains of APs. In 36 experiments under these conditions, AP trains induced e-
SP in 14 cases (39 %) and DSE in other 12 cases (33 %) (Figures 3A-3E). Both AP-evoked e-SP and DSE were abolished after perfusion with AM251 (n = 6 and 5, respectively) (Figures 3D and
3E). Furthermore, in some cases (n = 11), we first recorded the postsynaptic neuron in voltageclamp conditions to identify synapses displaying ND-evoked e-SP or DSE, and then we recorded the stimulating neuron in current-clamp conditions and applied depolarizing pulses. In these conditions, synapses displaying e-SP or DSE after ND also displayed the respective synaptic modulation after AP generation (n = 5 and 3 synapses, respectively). These results indicate that ND and APs similarly evoked e-SP and DSE.
We next determined the dependence of the e-SP and DSE on the neuronal activity level by applying depolarizing pulses of different durations that evoked different number of APs. Figure 3F shows that the degree of e-SP and DSE in previously identified synapses increased as the number of

7
APs increased. These results indicate that both synaptic phenomena can be evoked by more physiological stimuli (i.e., postsynaptic action potentials) and that they depend on the level of neuronal activity.
Because endocannabinoid signaling may be affected by the external temperature, we analyzed the e-SP at 24 ºC and 34 ºC. The ND-evoked potentiation of synaptic efficacy and Pr was similarly present at both temperatures (Figure S3), indicating that the ND-evoked e-SP was also present at a more physiological temperature.
Endocannabinoid-induced synaptic potentiation requires astrocyte calcium elevations
Because endocannabinoids elevate astrocyte Ca
2+
through CB1R activation (Navarrete and
Araque, 2008), and astrocytic Ca
2+
stimulates the release of gliotransmitters that modulate synaptic transmission (Araque et al., 1998; Haydon and Carmignoto, 2006; Perea and Araque, 2005, 2007;
Volterra and Meldolesi, 2005), we investigated if the e-SP required astrocyte Ca
2+
elevations. We first confirmed that endocannabinoids released by pyramidal neurons signal to adjacent astrocytes.
The astrocyte Ca
2+
signal, quantified from the Ca
2+
spike probability, was increased after ND or AP firing, and this effect was abolished by AM251 (cf. Navarrete and Araque, 2008) (Figures S2).
Then we tested whether astrocyte Ca
2+
was required to induce e-SP. Thapsigargin (1

M), which depletes the intracellular Ca 2+ stores by inhibiting the Ca 2+ ATPase (Navarrete and Araque, 2008;
Perea and Araque, 2005), abolished both the endocannabinoid-evoked astrocyte Ca
2+
elevations
(Navarrete and Araque, 2008) (Figures 4A-4C) and the e-SP (n = 7 cases that showed e-SP before thapsigargin perfusion; Figures 4D-4G). In contrast, the endocannabinoid-induced DSE observed in other 5 cases in control was unaffected (Figure S4). These results suggest that e-SP, but not DSE, requires astrocyte Ca 2+ elevations.
Because thapsigargin may not specifically act on astrocytes, to confirm this idea we specifically prevented Ca
2+
elevations in astrocytes by including the Ca
2+
chelator BAPTA (40 mM) in the astrocyte whole-cell recording pipette (Figures 5A-5D), because BAPTA dialyzed in a single

8 hippocampal astrocyte is known to spread to a large number of gap-junction connected astrocytes
(Jourdain et al., 2007; Serrano et al., 2006; Shigetomi et al., 2008). After confirming that NDevoked astrocyte Ca
2+
elevations were prevented in an area at least 150 µm around the recorded astrocyte (Figure 5C), we tested the consequences of BAPTA dialysis into the astrocyte network.
Because previous identification of synaptic modulatory effects in control resulted unfeasible, in these experiments we compared the incidence of DSE and e-SP (i.e., the proportion of synapses that showed either phenomenon) versus control slices. In 15 recordings under these conditions, ND induced DSE in 6 cases (i.e., 40 % vs. 27 % in control; P > 0.05), but failed to induce e-SP in any of the 15 cases (0 % vs.
27 % in control; P < 0.001) (Figures 5D, 5G and 5H).
We further confirmed the involvement of astrocytes in the synaptic modulation by including 10 mM GDPβS, which prevents G protein-mediated intracellular signalling, in the astrocyte whole-cell recording pipette (Figures 5E-5H). Like BAPTA, GDPβS dialyzed in a single hippocampal astrocyte could likely spread to a large number of gap-junction connected astrocytes, due to its relatively low molecular weight. We first confirmed that ND-evoked astrocyte Ca
2+
elevations were prevented in an area at least 150 µm around the recorded astrocyte (Figure 5E), and then we analyzed the consequences of GDPβS dialysis into the astrocyte network, comparing the incidence of DSE and e-SP versus control slices (because previous identification of synaptic modulatory phenomena in control was unfeasible). In 16 recordings under these conditions, ND induced DSE in
6 cases (i.e., 37 % vs . 27 % in control; P > 0.05), but failed to induce e-SP (only 1 out of 16 displayed e-SP, i.e., 6 % vs.
27 % in control; P < 0.001) (Figures 5F, 5G and 5H). Altogether, these results indicate that G protein-mediated astrocyte Ca
2+
elevations are not involved in DSE (cf.
Beierlein and Regehr, 2006) but they are required to induce e-SP.
Endocannabinoid-induced synaptic potentiation is mediated by metabotropic glutamate receptor type 1

9
Because astrocyte Ca 2+ elevations stimulate the release of glutamate which activating presynaptic metabotropic glutamate receptors (mGluRs) potentiates synaptic transmitter release
(Perea and Araque, 2007), we investigated if that mechanism was responsible for the e-SP. After identifying synapses that transiently increased Pr upon ND in control, we perfused the antagonists of group I mGluRs MPEP (50

M) and LY367385 (100

M; mGluR
5
and mGluR
1
antagonists, respectively). They blocked the e-SP (in control, Pr = 0.33 ± 0.05 and 0.55 ± 0.09 and in
MPEP+LY, Pr = 0.27 ± 0.05 and 0.29 ± 0.06, before and after ND, respectively; n = 7; not shown) without affecting the ND-evoked astrocyte Ca
2+
signal (not shown; cf. Navarrete and Araque,
2008). Likewise, e-SP evoked by postsynaptic AP generation was blocked by these mGluR antagonists (in control, Pr = 0.35 ± 0.02 and 0.55 ± 0.02, and in MPEP+LY, Pr = 0.33 ± 0.03 and
0.31 ± 0.02 before and after APs, respectively; n = 5). These results indicate that e-SP was mediated by activation of group I mGluRs.
We next investigated the subtype of mGluR involved by perfusing separately the group I mGluR antagonists (Figure 6). While LY367385 abolished the ND-evoked e-SP previously observed in control (Figures 6D-6G) without modifying the astrocyte Ca
2+
signal (Figure 6A-6C),
MPEP did not prevent e-SP (Figure 6H), indicating that e-SP was specifically mediated by activation of type 1 mGluRs.
To confirm that e-SP was mediated by glutamate released from astrocytes stimulated by endocannabinoids, and not by direct presynaptic CB1R activation, we used two experimental approaches. First, we monitored synaptic transmission and selectively increased intracellular Ca
2+ in astrocytes by whole-cell recording single astrocytes that were loaded with the Ca
2+
-cage NP-
EGTA (5 mM) through the recording pipette. UV-flash photolysis consistently elevated Ca 2+ in the recorded astrocyte and increased Pr (from 0.48 ± 0.11 to 0.65 ± 0.05; n = 8; P < 0.01) in 8 out of 14 recorded synapses (cf. Perea and Araque, 2007). In agreement with previous report (Perea and
Araque, 2007), this astrocyte-induced synaptic potentiation was still present after perfussing with

10
AM251, but was abolished by group I mGluR antagonists (n = 8 and n = 5 respectively; Figure S5).
Second, in the presence of AM251 to block CB1Rs, we directly applied the mGluR agonist DHPG while monitoring synaptic transmission properties. To prevent possible effects mediated by postsynaptic activation of mGluRs, recorded neurons were dialyzed with 2 mM GDPβS included in the recording pipette. Consistent with several reports (e.g., Faas et al., 2002; Chevaleyre and
Castillo, 2003), slow activation of mGluRs by perfusion with 50 µM DHPG depressed synaptic transmission (n = 5 synapses; see Figure S6). In contrast, a transient increase of Pr was observed in
8 out of 15 recorded synapses by fast local application of DHPG delivered by pressure pulses (0.5 s) through a micropipette (Figure S6; in the other 3 and 4 cases out of the 15 recorded synapses Pr was decreased or unchanged, respectively). This effect was abolished after perfusion with group I mGluR antagonists (n = 5; Figures S6C-S6F), but no with LY367385 alone (n = 4; not shown), indicating that the synaptic potentiation evoked by exogenous glutamate receptor agonists involved mGluR1 and mGluR5, whereas e-SP induced by astrocytic glutamate selectively involved mGluR1
(see Fig. 6). Differences in the effects of DHPG observed by delivery through perfusion or pressure pulses may be accounted for by the different application kinetics that may lead to inhibition or facilitation of transmitter release (Rodriguez-Moreno et al., 1998; Sanchez-Prieto et al., 2004). In summary, e-SP was abolished by mGluR antagonists, Ca
2+
elevations in astrocytes induced CB1Rinsensitive and mGluR-dependent increase of transmitter release, and mGluR activation could transiently enhance transmitter release. Therefore, these results indicate that e-SP is mediated by mGluR activation (Figure 6I).
While group I mGluRs are coupled to pertussis toxin (PTX)-insensitive G q/11 proteins (Conn and
Pin, 1997; Miura et al., 2002), CB1Rs in neurons are mainly coupled to PTX-sensitive G i/o
proteins
(Chevaleyre et al., 2006; Mackie and Hille, 1992; Piomelli, 2003). We then investigated the nature of the G protein involved by preincubating the slices with 7.5 μg/ml PTX for 3.5-9 h. Because the long incubation time prevented the previous identification of synapses in control, we compared the

11 incidence of DSE and e-SP in parallel control and PTX treated slices. In 9 PTX treated slices, only
2 out of 21 (9 %) synapses showed DSE ( vs . 6 out of 16 synapses, 37 %, in 6 parallel untreated slices; P < 0.001). In contrast, ND-evoked e-SP occurred in 12 out of 21 synapses from 9 PTXtreated slices (57 % vs.
33 % in 6 parallel control slices; P < 0.01) (Figures 7A and 7B). These results indicate that 1) G i/o
proteins mediate the CB1R-induced DSE, 2) e-SP is mediated by activation of type I mGluRs coupled to a PTX-insensitive G protein; and 3) e-SP requires PTXinsensitive CB1R-mediated signaling. Furthermore, the higher proportion of synapses that showed e-SP in PTX-treated slices compared to control indicates that more synapses were potentiated after blocking the CB1R-mediated intracellular signaling, suggesting that in control conditions this signaling prevails over the mGluR-mediated signaling (see below).
The sign of the endocannabinoid-induced synaptic modulation does not depend on differential expression of CB1Rs but on specific intercellular signaling pathways
To investigate if the sign of the endocannabinoid-induced synaptic modulation (i.e., DSE or e-
SP) was due to differential expression of CB1Rs by synaptic terminals or to specific intercellular signaling pathways, we asked whether synapses that showed e-SP could be depressed by activation of CB1Rs. We performed control recordings, and after assessing that the recorded synapse showed e-SP, we perfused the CB1R agonist WIN55-212,2 (5 µM). In all the 7 synapses that displayed e-
SP, WIN application depressed synaptic efficacy (Figures 7C-7E). Although we cannot exclude that
WIN effects on basal synaptic transmission might be due to increasing action potential conduction failures, the fact that ND-evoked e-SP was prevented by WIN (Fig. 7C-7E) indicates that synapses that showed e-SP also expressed CB1Rs linked to intracellular cascades that can lead to synaptic depression, but these receptors were probably spared from the released endocannabinoids.
Furthermore, this result along with the facts that WIN-treated synapses no longer showed e-SP after
ND (Figures 7C-7E) and that the incidence of e-SP was higher in PTX-treated slices (see above)

12 indicate that direct activation of presynaptic CB1Rs, which leads to DSE, overpowers the mGluRmediated signaling that leads to e-SP.
Endocannabinoids induce homoneuronal synaptic depression and heteroneuronal synaptic potentiation
While endocannabinoids released by neurons exert their effects at short-distances (< 20 µm;
Chevaleyre et al., 2006; Chevaleyre and Castillo, 2004; Piomelli, 2003; Wilson and Nicoll, 2001), the astrocyte Ca
2+
signal, which spreads along the cell, may have long-distance neuromodulatory effects by releasing gliotransmitters at distal regions (Perea and Araque, 2005; Serrano et al., 2006).
We therefore hypothesized that endocannabinoids released by ND depress synaptic transmission acting directly on CB1Rs at near synaptic terminals, whereas they potentiate more distant synapses through activation of astrocytic CB1Rs, which leads to Ca 2+ elevations in astrocytes that spread intracellularly and stimulate glutamate release at distant sites. To test this hypothesis we investigated the spatial profile of the endocannabinoid-mediated synaptic modulation, i.e. the appearance of DSE and e-SP in synapses as a function of distance from the neighboring stimulated neuron (Figure 8A; because the exact location of synapses could not be determined, the distance of synapses was estimated from the distance of the somas of the stimulating and recording neurons).
While most of the synapses displaying DSE (i.e., Pr < 100% from basal values) were located at relatively short distances from the stimulated neuron, those undergoing e-SP (i.e., Pr > 100%) were located at relatively longer distances.
To investigate the possible functional consequences of the spatial segregation of both synaptic modulatory phenomena, we analyzed the simultaneous effects of endocannabinoids on the synaptic properties of synapses in the depolarized neuron (homoneuronal synapses) and in the nonstimulated neuron (heteroneuronal synapses) (Figures 8B-8E). In 33 paired recorded synapses, ND induced DSE in 13 homoneuronal synapses (39 %; Figure S7), but e-SP was never observed (0 %).

13
Contrastingly, in heteroneuronal synapses, 10 showed DSE (30 %) and 9 displayed e-SP (27 %; n =
33; Figures 8B-8E).
The absence of e-SP in homoneuronal synapses is consistent with the short range direct effect of endocannabinoids and the prevalence of CB1R- over mGluR-mediated signaling. These ideas predict that e-SP in homoneuronal synapses would be canceled by the CB1R-mediated signaling but could be unmasked by blocking this signaling cascade. To test this hypothesis, we analyzed the incidence of e-SP in homoneuronal synapses from 9 slices treated with PTX. In these slices, DSE occurred in only 3 out of 21 homoneuronal synapses ( vs . 13 out of 33 in control slices; i.e., 14 % vs.
39 %; P < 0.001), confirming that the CB1R-mediated signaling was blocked. In this condition, e-
SP, which was absent in 33 control homoneuronal synapses, was present in 4 out of 13 homoneuronal synapses (i.e., 31 % in PTX-treated slices vs.
0 % in control slices; P < 0.001).
Taken together, these results indicate that synaptic modulation by endocannabinoids is spatiallycontrolled and result from activation of specific intracellular signals, i.e., short-range direct activation of presynaptic CB1Rs or long-range indirect activation of mGluRs through stimulation of astrocytes (Figure 8F).
DISCUSSION
Present results indicate that endocannabinoids may have opposite neuromodulatory effects, depressing synaptic transmission by directly acting on presynaptic CB1Rs or potentiating synaptic transmission by acting on astrocytic CB1Rs that stimulate glutamate release from astrocytes. While endocannabinoids are known to exert their direct inhibitory effects at short-distances (< 20 µm; see
Chevaleyre et al., 2006; Chevaleyre and Castillo, 2004; Piomelli, 2003; Wilson and Nicoll, 2001), our results suggest that the effects of endocannabinoid signaling are spatially prolonged by the endocannabinoid-induced astrocyte Ca
2+
signal, which spreading along the cell may have longdistance neuromodulatory effects by releasing gliotransmitters at distal regions (Perea and Araque,
2005; Serrano et al., 2006). Therefore, endocannabinoids represent a short-range signal that lead to

14 direct depression of synapses close to the endocannabinoid source, but by stimulating astrocytes they trigger a long-range signal (i.e., the astrocyte intracellular Ca
2+
that triggers glutamate release) that indirectly lead to synaptic potentiation of relatively more distant synapses.
While CB1R activation by endocannabinoids is well known to induce synaptic depression in many brain regions (Chevaleyre et al., 2006; Kano et al., 2009; Heifets and Castillo, 2009), present data indicate that endocannabinoids may lead to the potentiation of hippocampal synaptic transmission. This synaptic potentiation is an indirect effect mediated by the stimulation of astrocytes that then release the gliotransmitter glutamate. Likewise, it has been recently reported that endocannabinoids may stimulate the release of dopamine that lead to the modulation of chemical and electrical synapses in the Mauthner cell of goldfish (Cachope et al., 2007). Whether indirect synaptic modulation induced by endocannabinoids through the stimulation of a second neuromodulator is a general phenomenon in the nervous system is an exciting possibility that needs to be investigated. Furthermore, the possible involvement of glia-mediated mechanisms in the endocannabinoid-induced modulation of neural network activity, such as the recently reported endocannabinoid-mediated effects on the spinal locomotor activity of the lamprey (Kyriakatos and
El Manira, 2007), requires further investigation.
The heterogeneity of endocannabinoid-induced synaptic modulation depends on spatially controlled intercellular signaling pathways
Using the minimal stimulation technique, which stimulates single or very few presynaptic fibers, we have found that endocannabinoids released from pyramidal neurons induced DSE, e-SP or no changes in synaptic transmission in different synapses. The variety of the synaptic modulation does not imply heterogeneity of the synapses but is probably a consequence of the experimental conditions, where single or very few presynaptic fibers were stimulated and a single pyramidal neuron was recorded. Indeed, the different synaptic changes observed in specific synapses cannot be accounted for by different intrinsic properties of synaptic terminals (e.g., differential expression

15 of CB1 or mGluRs) because they can undergo depression or potentiation upon exogenous activation of these receptors. Rather, they depend on specific intercellular signaling pathways, whose activation is spatially controlled (Figure 8A). Indeed, either direct activation of CB1Rs or indirect activation of mGluRs through stimulation of astrocytes depends on the distance between the synapse and the source of endocannabinoids.
Consequently, synapses located at relatively long distance of the endocannabinoid source did not show DSE because CB1Rs could not be activated due to the short range of action of endocannabinoids (Chevaleyre and Castillo, 2004; Piomelli, 2003; Wilson and Nicoll, 2001), but they displayed mGluR-mediated e-SP because astrocytes activated by endocannabinoids could expand the signaling, through the extension of the intracellular, and perhaps also intercellular, Ca
2+ and the subsequent glutamate release at relatively distant regions (Volterra and Melodelsi, 2005;
Haydon and Carmignoto, 2006; Perea et al., 2009). In contrast, DSE was favored in synapses located at relatively short distance of the endocannabinoid source because the signaling cascade triggered by CB1R activation prevails over that stimulated by mGluR activation (Figure 7).
Although we observed DSE in neurons whose somas were located between 20 and 60 µm away from the stimulated neuron, these results are not necessarily inconsistent with short-range (< 20
µm) endocannabinoid signaling (Chevaleyre et al., 2006; Chevaleyre and Castillo, 2004; Piomelli,
2003; Wilson and Nicoll, 2001) because, due to the dendritic arborization, the modulated synapses may be closer to the stimulated neuron than the respective neuronal soma.
The spatially controlled sign of the endocannabinoid-induced synaptic modulation may have important functional implications. Furthermore, regulatory mechanisms that regulate the spatial extension of endocannabinoid signaling, such as diffusion or degradation of endocannabinoids, may have relevant consequences because they can regulate the appearance of DSE or e-SP.
Subcellular localization of receptors

16
While DSE is known to be mediated by activation of CB1Rs located at presynaptic terminals
(see e.g., Alger, 2002; Chevaleyre et al., 2006), present results indicate that e-SP is indirectly mediated by activation of CB1Rs located in astrocytes. This idea is supported by the fact that astrocytes express functional CB1Rs (Stella, 2004; Navarrete and Araque, 2008) that upon activation elevate the intracellular Ca
2+
and stimulates glutamate release (Navarrete and Araque,
2008), and that e-SP is mediated by mGluRs and requires both CB1R activation and astrocytic Ca 2+ elevations (Figures 4-6).
Astrocytic CB1R activation stimulates glutamate release that then activates group I mGluRs to induce e-SP. Several evidence indicate that these mGluRs are located presynaptically: 1) e-SP is characterized by the enhancement of transmitter release probability without changes in synaptic potency, which indicates a presynaptic rather than a postsynaptic underlying mechanism; 2) e-SP is accompanied by changes in the paired-pulse facilitation (Figures S1A and S1B), which is consistent with a presynaptic mechanism; 3) a transient enhancement of transmitter release probability could be induced by application of the mGluR agonist DHPG when the postsynaptic signaling cascade of mGluRs was blocked by dialyzing the postsynaptic neuron with GDPβS. Taken together, these results strongly indicate that e-SP is mediated by activation of presynaptic mGluRs. However, anatomical evidence for presynaptic mGluR1 at Schaffer collaterals awaits confirmation.
Physiological consequences of endocannabinoid-mediated synaptic modulation
In contrast to the large amount of evidence indicating the relevance of the astrocyte Ca 2+ signal in neurophysiology (for reviews see Haydon and Carmignoto, 2006; Perea et al., 2009), recent reports have questioned its physiological importance (Agulhon et al., 2010). Present data demonstrate that astrocyte Ca
2+
signal triggered by an endogenous ligand (i.e. endocannabinoids) released by physiological events (i.e. neuronal action potentials) is competent to modulate synaptic transmission properties.

17
Present results indicate that astrocytes exert a subtle modulatory effect on the activity of synapses that was overlooked when analyzing the effects on neurotransmission at large scale.
Indeed, the fact that homoneuronal synapses only showed DSE suggests that the endocannabinoidinduced synaptic depression mediated by direct activation of presynaptic CB1Rs is the prevailing mechanism over the astrocyte-induced synaptic potentiation. Consequently, it is not surprising that endocannabinoid-induced synaptic potentiation remained elusive when synaptic transmission was analyzed by massive stimulation of axons or when exogenous agonists were applied. Therefore, our results widen our view of the delicate regulatory role of astrocytes in neurotransmission, which may be concealed by stronger modulatory mechanisms, but which can be revealed by analyzing synaptic transmission in fine detail.
In conclusion, astrocytes potentiate hippocampal synaptic efficacy upon stimulation by endocannabinoids, indicating that they are an integral part of the endocannabinoid system. These results expand our view of endocannabinoid signaling as well as astrocytic function on synaptic physiology, and indicate that astrocytes are involved in relevant phenomena of brain function.

18
EXPERIMENTAL PROCEDURES
Hippocampal slice preparation
Hippocampal slices were obtained from C57BL/6 mice (13-18 days old). In some cases, slices from CB1 receptor knock-out mice, generously donated by Dr. A. Zimmer, were used (Zimmer et al., 1999). All the procedures for handling and sacrificing animals followed the European
Commission guidelines (86/609/CEE). Animals were anaesthetized and decapitated. The brain was rapidly removed and placed in ice-cold artificial cerebrospinal fluid (ACSF). Slices (350-400 μm thick) were incubated during >1 h at room temperature (21-24º C) in ACSF that contained (in mM):
NaCl 124, KCl 2.69, KH
2
PO
4
1.25, MgSO
4
2, NaHCO
3
26, CaCl
2
2 and glucose 10, and was gassed with 95% O
2
/ 5% CO
2
(pH = 7.3). Slices were then transferred to an immersion recording chamber and superfused at 2 ml/min (the chamber volume was replaced in 10-12 min) with gassed ACSF including 0.05 mM Picrotoxin and 100 µM Saclofen to block GABA
A
and GABA
B
receptors respectively. Cells were visualized under an Olympus BX50WI microscope (Olympus Optical,
Tokyo, Japan) equipped with infrared and differential interference contrast imaging devices, and with a 40x water immersion objective.
Electrophysiology
Simultaneous electrophysiological recordings from two CA1 pyramidal neurons were made using the whole-cell patch-clamp technique. Patch electrodes had resistances of 3-10 MΩ when filled with the internal solution that contained (in mM) for pyramidal neurons: KGluconate 135,
KCl 10, HEPES 10, MgCl 1, ATP-Na
2
2 (pH = 7.3). Recordings were obtained with PC-ONE amplifiers (Dagan Instruments, Minneapolis, MN). Fast and slow whole-cell capacitances were neutralized and series resistance was compensated (≈ 70 %), and the membrane potential was held at -70 mV. Electrophysiological properties were monitored before and at the end of the experiments. Series and input resistances were monitored throughout the experiment using a -5 mV pulse. Recordings were considered stable when the series and input resistances, resting membrane

19 potential and stimulus artefact duration did not change > 20%. Furthermore, I-V curves and firing pattern before and at the end of the experiments were similar. Cells that did not meet these criteria were discarded. Signals were fed to a Pentium-based PC through a DigiData 1440A interface board.
Signals were filtered at 1 KHz and acquired at 10 KHz sampling rate. The pCLAMP 10.2 (Axon instruments) software was used for stimulus generation, data display, acquisition and storage. We performed paired whole-cell recordings from two CA1 pyramidal neurons (distance of the somas <
150 μm). In some experiments astrocytes were patched with 4-9 MΩ electrodes filled with an intracellular solution containing (in mM): 1 MgCl
2
, 8 NaCl, 2 ATP, 0.4 GTP, 10 HEPES, and either
40 mM BAPTA or 10 mM GDPβS, titrated with KOH to pH 7.2–7.25 and adjusted to 275–285 mOsm. The Ca
2+
chelator BAPTA and GDPβS were included to prevent Ca
2+
elevations and G protein-mediated intracellular signaling, respectively, in astrocytes. Astrocyte whole-cell recordings lasted at least 30 min to allow the dialysis of BAPTA or GDPβS throughout the gap-junction connected astrocyte network. We performed control experiments similar to regular experiments but placing the BAPTA containing pipette in the bath without achieving whole-cell mode. After 1-2 minutes of positive pressure, the pipette was withdrawn. The subsequent analysis of synaptic transmission properties indicate that both DSE and e-SP could be evoked by ND (n = 12; 33 % and
33 % respectively).
In some cases, the stimulating neuron was recorded in current-clamp conditions and depolarizing pulses of different durations were applied to evoke trains of action potentials. When long duration pulses induced strong neuronal firing adaptation, shorter repetitive pulses were delivered. The duration of the depolarizing pulses was varied to obtain different number of action potentials.
Synaptic stimulation
Theta capillaries (2–5 μm tip) filled with ACSF was used for bipolar minimal stimulation. The electrodes were connected to a stimulator S-910 through an isolation unit and placed in the stratum

20 radiatum to stimulate Schaffer collateral (SC) afferents. Single pulses (250 μs duration) or paired pulses (50 ms interval) were continuously delivered at 0.33 Hz. The stimulus intensity (5-10 mA) was adjusted to meet the conditions that putatively stimulate single or very few presynaptic fibers
(c.f. Dobrunz and Stevens, 1997; Isaac et al., 1996; Perea and Araque, 2007; Raastad, 1995), and was unchanged during the experiment. Recordings that did not meet these criteria and synapses that did not show amplitude stability of EPSCs were rejected. The synaptic parameters analyzed were: synaptic efficacy (mean peak amplitude of all responses including failures); synaptic potency (mean peak amplitude of the successes); probability of release (Pr, ratio between number of successes versus total number of stimuli); and paired-pulse facilitation (PPF = [(2nd EPSC – 1st EPSC)/ 1st
EPSC]). Synaptic parameters were determined from 50 stimuli before (basal) and after stimulus
(ND or APs). Two consecutive responses to ND were averaged. To illustrate the time course of
ND-induced effects, synaptic parameters were grouped in 33 s bins. EPSC amplitude was determined as the peak current amplitude (2-9 ms after stimulus) minus the mean baseline current
(20-30 ms before stimulus). A response was considered a failure if the amplitude of the current was
< 3 times the standard deviation of the baseline current (0.7 - 2.3 pA) and was verified by visual inspection.
The presence of e-SP or DSE was assessed in individual synapses if statistical significant differences were found between Pr values obtained 96 s before and after ND or APs. To characterize pharmacologically both phenomena, we first identified specific synapses undergoing either DSE or e-SP in control conditions and then analyzed the effects of subsequent superfusion of pharmacological agents on the synapses previously identified. Because in experiments involving
CB1R knock out mice, BAPTA-loaded cells, or PTX treatment, the previous identification of synapses in control was unfeasible, we compared the incidence of DSE and e-SP (i.e., the proportion of synapses that showed either phenomenon) in parallel control and test slices.

21
To asses that effects observed after PTX treatment were specifically due to the toxin, we tested the effects of slice preincubation with 7.5 μg/ml heat-inactivated PTX for 3.5-9 h. In 9 synapses recorded from 4 heat-inactivated-PTX treated slices and in 21 synapses from 6 control slices, the incidence of DSE and e-SP was similar (33 % and 37 % for DSE, and 33 % and 33 % for e-SP, in control and heat-inactivated PTX, respectively; P > 0.05).
Experiments were performed at room temperature (21-24 °C), unless stated otherwise. In some cases, temperature of the bath solution was adjusted and controlled to 24 ºC and 34 ºC with a temperature controller TC-324B (Warner Instruments Co.), using a probe immersed in the recording chamber. In some cases the temperature was elevated from 24 to 34 ºC, and in other cases it was decreased from 34 to 24 ºC. We then compared the relative effects of ND on synaptic transmission properties at both temperatures on the same synapses. Evidence of the effective change of temperature is provided by the relatively higher basal Pr at 34 ºC than at 24 ºC (see Figure S3) as well as by the faster EPSC decay time course (9.1 ± 0.4 and 6.2 ± 0.2 ms at 34 ºC and 24 ºC, respectively; n = 59; P < 0.001).
The effects of pharmacological agents were tested at < 40 min after entering whole-cell mode in the stimulating neuron, i.e., before a significant rundown of the ND-evoked effect took place (see
Figure S1E). Data are expressed as mean ± SEM. Results were compared using a two-tailed
Student’s t -test (α = 0.05), or the Chi-square test when comparing the proportion of DSE and e-SP.
Statistical differences were established with P < 0.05 (*), P < 0.01 (**), and P < 0.001 (***).
Ca 2+ imaging
Ca
2+ levels in astrocytes located in the stratum radiatum of the CA1 region of the hippocampus were monitored by fluorescence microscopy using the Ca
2+
indicator fluo-4 (Molecular Probes,
Eugene, OR). Slices were incubated with fluo-4-AM (2-5
 l of 2 mM dye were dropped over the hippocampus, attaining a final concentration of 2-10

M and 0.01 % of pluronic) for 20-30 min at room temperature. In these conditions, most of the cells loaded were astrocytes (Araque, et al.,

22
2002; Kang et al., 1998; Nett, et al., 2002; Parri et al., 2001; Perea and Araque, 2005) as confirmed in some cases by their electrophysiological properties. Astrocytes were imaged using a CCD camera (Retiga EX; Qimaging, Canada) attached to the Olympus microscope. Cells were illuminated during 200-500 ms with a xenon lamp at 490 nm using a monochromator Polychrome
II (T.I.L.L. Photonics, Planegg, Germany), and images were acquired every 0.5-1 s. The monochromator Polychrome II and the CCD camera were controlled and synchronized by the IP
Lab software (BD Biosciences, MD) that was also used for quantitative epifluorescence measurements. Intracellular Ca
2+
signal of stratum radiatum astrocytes was monitored from cells located within a 150 µm wide region around the stimulated neuron and perpendicular to the stratum pyramidale. Ca
2+
variations recorded at the soma of the cells were estimated as changes of the fluorescence signal over baseline (

F/F
0
), and cells were considered to respond to the stimulation when

F/F
0
increased three times the standard deviation of the baseline for at least two consecutive images and with a delay

15 s after the stimulation.
The astrocyte Ca
2+
signal was quantified from the probability of occurrence of a Ca
2+
spike and the Ca
2+
oscillation frequency. The Ca
2+
spike probability was calculated from the number of Ca
2+ elevations grouped in 5 s bins recorded from 6 to 14 astrocytes in the field of view. The time of occurrence was considered at the onset of the Ca
2+
spike. The Ca
2+
oscillation frequency was obtained from the number of Ca 2+ spikes occurring in 6 to 14 astrocytes in the field of view during
50 second periods before (basal) and after the onset of the neuronal depolarization. To test the effects of ND or APs on Ca
2+
spike probability under different conditions, the respective mean basal
(50 s before ND) and maximum Ca
2+
spike probability (i.e., 5-10 s after ND) from different slices were averaged and compared. Mean values were obtained from at least 4 slices in each condition.
Successive ND consistently evoked successive increases of the astrocyte Ca 2+ spike probability (see
Navarrete and Araque, 2008).
Ca 2+ uncaging by UV-flash photolysis.

23
In photo-stimulation experiments, single astrocytes were electrophysiologically recorded with patch pipettes filled with the internal solution containing 5 mM NP-EGTA (and 50

M fluo-4 to monitor Ca
2+
levels). Ca
2+
uncaging was achieved by delivering train pulses (1 ms duration, 6-15 mW) of UV light (340-380 nm) at 2 Hz during 5 s to the soma and processes of the recorded astrocyte (optical window of 15-25
 m diameter) using a flash photolysis system (Rapp
OptoElectronic, Hamburg, Germany)..
Drugs and chemicals
1,2-bis(2-aminophenoxy)ethaneN , N , N
′,
N
′-tetraacetate
(BAPTA); N-(Piperidin-1-yl)-5-(4iodophenyl)-1-(2,4-dichlorophenyl)-4-methyl-1H-pyrazole-3-carboxamide (AM-251); 2-Methyl-6-
(phenylethynyl)pyridine hydrochloride (MPEP), (S)-(+)-α-Amino-4-carboxy-2methylbenzeneacetic acid (LY367385), (RS)-3-Amino-2-(4-chlorophenyl)propylsulfonic acid
(saclofen), thapsigargin and (RS)-3,5-DHPG were purchased from Tocris Cookson (Bristol, UK), and Fluo-4-AM and o-nitrophenyl EGTA, tetrapotassium salt (NP-EGTA) (Molecular Probes,
Eugene, OR). All other drugs were from Sigma.

24
REFERENCES
Alger, B.E. (2002). Retrograde signaling in the regulation of synaptic transmission: focus on endocannabinoids. Prog. Neurobiol. 68, 247-2 86.
Agulhon, C., Fiacco, T.A., and McCarthy, K.D. (2010) Hippocampal short- and long-term plasticity are not modulated by astrocyte Ca
2+
signaling. Science 327 , 1250-1254.
Araque, A., Carmignoto, G., and Haydon, P.G. (2001). Dynamic signaling between neurons and glia. Annu. Rev. Physiol. 63, 795-813.
Araque, A., Martin, E.D., Perea, G., Arellano, J.I., and Buño, W. (2002). Synaptically released acetylcholine evokes Ca
2+
elevations in astrocytes in hippocampal slices. J. Neurosci. 22, 2443-
2450.
Araque, A., Parpura, V., Sanzgiri, R.P., and Haydon, P.G. (1998). Glutamate-dependent astrocyte modulation of synaptic transmission between cultured hippocampal neurons. Eur. J. Neurosci. 10,
2129-2142.
Araque, A., Parpura, V., Sanzgiri, R.P., and Haydon, P.G. (1999). Tripartite synapses: glia, the unacknowledged partner. Trends. Neurosci. 22, 208-215.
Beattie, E.C., Stellwagen, D., Morishita, W., Bresnahan, J.C., Ha, B.K., Von Zastrow, M., Beattie,
M.S., and Malenka, R.C. (2002). Control of synaptic strength by glial TNF

. Science 295, 2282-
2285.
Beierlein, M., and Regehr, W.G. (2006) Brief bursts of parallel fiber activity trigger calcium signals in Bergmann glia. J. Neurosci. 26 , 6958-6967.
Cachope, R., Mackie, K., Triller, A., O'Brien, J., and Pereda, A.E. (2007). Potentiation of electrical and chemical synaptic transmission mediated by endocannabinoids. Neuron 56, 1034-1047.
Chevaleyre, V., and Castillo, P.E. (2003). Heterosynaptic LTD of hippocampal GABAergic synapses: a novel role of endocannabinoids in regulating excitability. Neuron 38, 461-472
.
Chevaleyre, V., and Castillo, P.E. (2004). Endocannabinoid-mediated metaplasticity in the hippocampus. Neuron 43, 871-881.
Chevaleyre, V., Takahashi, K.A., and Castillo, P.E. (2006). Endocannabinoid-mediated synaptic plasticity in the CNS. Annu. Rev. Neurosci. 29, 37-76.
Conn, P.J., and Pin, J.P. (1997). Pharmacology and functions of metabotropic glutamate receptors.
Annu. Rev. Pharmacol. Toxicol. 37, 205-237.
Di Marzo, V., Fontanta, A., Cadas, H., Schinelli, S., Cimino, G., Schwartz, J.C., and Piomelli, D.
(1994). Formation and inactivation of endogenous cannabinoid anandamide in central neurons.
Nature 372, 686-691.
Dobrunz, L.E., and Stevens, C.F. (1997). Heterogeneity of release probability, facilitation, and depletion at central synapses. Neuron 18, 995-1008.

25
Faas, G.C., Adwanikar, H., Gereau, R.W. 4 th ., and Saggau, P. (2002). Modulation of presynaptic calcium transients by metabotropic glutamate receptor activation: A differential role in acute depression of synaptic transmission and long-term depression. J. Neurosci. 22, 6885–6890.
Freund, T.F., Katona, I., and Piomelli, D. (2003). Role of endogenous cannabinoids in synaptic signaling. Physiol. Rev. 83, 1017-1066.
Hashimotodani, Y., Ohno-Shosaku, T., and Kano, M. (2007). Endocannabinoids and synaptic function in the CNS. Neuroscientist. 13, 127-137.
Haydon, P.G., and Carmignoto, G. (2006). Astrocyte control of synaptic transmission and neurovascular coupling. Physiol. Rev. 86, 1009-10031.
Heifets, B.D., and Castillo, P.E. (2009). Endocannabinoid Signaling and Long-Term Synaptic
Plasticity Annu. Rev. Physiol. 71, 283–306.
Isaac, J.T., Hjelmstad, G.O., Nicoll, R.A., and Malenka, R.C. (1996). Long-term potentiation at single fiber inputs to hippocampal CA1 pyramidal cells. Proc. Natl. Acad. Sci. USA. 93, 8710-
8715.
Jourdain, P., Bergersen, L.H., Bhaukaurally, K., Bezzi, P., Santello, M., Domercq, M., Matute, C.,
Tonello, F., Gundersen, V., and Volterra, A. (2007). Glutamate exocytosis from astrocytes controls synaptic strength. Nat. Neurosci. 10, 331-339.
Kang, J., Jiang, L., Goldman, S.A., and Nedergaard, M. (1998) Astrocyte-mediated potentiation of inhibitory synaptic transmission. Nat. Neurosci. 1, 683-692.
Kano, M., Ohno-Shosaku, T., Hashimotodani, Y., Uchigashima, M., and Watanabe, M. (2009).
Endocannabinoid-mediated control of synaptic transmission. Physiol Rev. 89, 309-380.
Kreitzer, A.C., and Regehr, W.G. (2001). Retrograde inhibition of presynaptic calcium influx by endogenous cannabinoids at excitatory synapses onto Purkinje cells. Neuron 29, 717-727.
Kreitzer, A.C., and Regehr, W.G. (2002). Retrograde signaling by endocannabinoids. Curr. Opin.
Neurobiol. 12, 324-330.
Kyriakatos, A., and El Manira, A. (2007) Long-term plasticity of the spinal locomotor circuitry mediated by endocannabinoid and nitric oxide signaling. J. Neurosci. 27 , 12664-12674.
Lovinger, D.M. (2008). Presynaptic modulation by endocannabinoids. Handb. Exp. Pharmacol.
184, 435-477.
Mackie, K., and Hille, B. (1992). Cannabinoids inhibit N-type calcium channels in neuroblastomaglioma cells. Proc. Natl. Acad. Sci. USA 89, 3825-3829.
Maldonado, R., Valverde, O., and Berrendero, F. (2006). Involvement of the endocannabinoid system in drug addiction. Trends Neurosci. 29, 225-232.
Miura, M., Watanabe, M., Offermanns, S., Simon, M.I., and Kano, M. (2002). Group I metabotropic glutamate receptor signaling via Galpha q/Galpha 11 secures the induction of longterm potentiation in the hippocampal area CA1. J. Neurosci. 22, 8379-8390.

26
Navarrete, M., and Araque, A. (2008). Endocannabinoids mediate neuron-astrocyte communication.
Neuron 58, 883-893.
Nedergaard, M., Ransom, B., and Goldman, S.A. (2003). New roles for astrocytes: Redefining the functional architecture of the brain. Trends Neurosci. 26, 523-530.
Nett, W.J., Oloff, S.H., and McCarthy, K.D. (2002). Hippocampal astrocytes in situ exhibit calcium oscillations that occur independent of neuronal activity. J. Neurophysiol. 87, 528-537.
Ohno-Shosaku, T., Maejima, T., and Kano, M. (2001). Endogenous cannabinoids mediate retrograde signals from depolarized postsynaptic neurons to presynaptic terminals. Neuron 29, 729-
738.
Ohno-Shosaku, T., Tsubokawa, H., Mizushima, I., Yoneda, N., Zimmer, A., and Kano, M. (2002).
Presynaptic cannabinoid sensitivity is a major determinant of depolarization-induced retrograde suppression at hippocampal synapses. J. Neurosci. 22, 3864-3872.
Parri, H.R., Gould, T.M., and Crunelli, V. (2001). Spontaneous astrocytic Ca
2+
oscillations in situ drive NMDAR-mediated neuronal excitation. Nat. Neurosci. 4, 803-812.
Perea, G., and Araque, A. (2005). Properties of synaptically evoked astrocyte calcium signal reveal synaptic information processing by astrocytes. J. Neurosci. 25, 2192-2203.
Perea, G., and Araque, A. (2007). Astrocytes potentiate transmitter release at single hippocampal synapses. Science 317, 1083-1086.
Perea, G., Navarrete, M., and Araque, A. (2009). Tripartite synapses: astrocytes process and control synaptic information. Trends. Neurosci. 32, 421-231.
Piomelli, D. (2003). The molecular logic of endocannabinoid signaling. Nat. Rev. Neurosci. 4, 873-
884.
Porter, J.T., and McCarthy, K.D. (1996). Hippocampal astrocytes in situ respond to glutamate released from synaptic terminals. J. Neurosci. 16, 5073- 5081.
Raastad, M. (1995). Extracellular activation of unitary excitatory synapses between hippocampal
CA3 and CA1 pyramidal cells. Eur J Neurosci. 7 , 1882-1888.
Rodríguez-Moreno, A., Sistiaga, A., Lerma, J., Sánchez-Prieto, J. (1998). Switch from facilitation to inhibition of excitatory synaptic transmission by group I mGluR desensitization. Neuron. 21 ,
1477-1486.
Sánchez-Prieto, J., Paternain, A.V., and Lerma , J. (2004) Dual signaling by mGluR5a results in bidirectional modulation of N-type Ca 2+ channels. FEBS Lett.. 57, 428-432.
Serrano, A., Haddjeri, N., Lacaille, J.C., and Robitaille, R. (2006). GABAergic network activation of glial cells underlies hippocampal heterosynaptic depression. J. Neurosci. 26, 5370-5382.
Shigetomi, E., Bowser, D.N., Sofroniew, M.V., and Khakh, B.S. (2008). Two forms of astrocyte calcium excitability have distinct effects on NMDA receptor-mediated slow inward currents in pyramidal neurons. J. Neurosci. 28, 6659-6663.

27
Volterra, A., and Bezzi, P. (2002). The Tripartite Synapse: Glia In Synaptic Transmission eds.
Volterra, A., Magistretti, P.J., and Haydon, P.G. (New York: Oxford University. Press), pp. 164-
184.
Volterra, A., and Meldolesi, J. (2005). Astrocytes, from brain glue to communication elements: the revolution continues. Nat. Rev. Neurosci. 6, 626-640.
Wilson, R.I., and Nicoll, R.A. (2001). Endogenous cannabinoids mediate retrograde signaling at hippocampal synapses. Nature 410, 588-592.
Wilson, R.I., and Nicoll, R.A. (2002). Endocannabinoid signaling in the brain. Science 296, 678-
682.
Zimmer, A., Zimmer, A.M., Hohmann, A.G., Herkenham, M., and Bonner, T.I. (1999). Increased mortality, hypoactivity, and hypoalgesia in cannabinoid CB1 receptor knockout mice. Proc. Natl.
Acad. Sci.USA 96, 5780-5785.

28
FIGURE LEGENDS
Figure 1. Endocannabinoids potentiate transmitter release at hippocampal CA3-CA1 synapses.
(A) Infrared differential interference contrast (DIC) image showing paired recorded CA1 pyramidal neurons and the stimulation electrode (bottom). Scale bar, 40
 m. (B) Schematic drawing depicting paired recordings from pyramidal neurons and the stimulating electrode. ND, neuronal depolarization. (C) Responses evoked by minimal stimulation showing regular EPSC amplitudes and failures of synaptic transmission (20 consecutive stimuli) (top traces) and averaged EPSCs (n =
50 stimuli, including successes and failures; bottom traces) before (basal) and after ND in control and after perfusion with 2 μM AM251. (D) EPSC amplitudes from a representative synapse in control and AM251. Zero time corresponds to the beginning of ND, as in E and all other figures.
(E) Synaptic efficacy (i.e., mean amplitude of responses including successes and failures of neurotransmission), probability of neurotransmitter release (Pr) and synaptic potency (i.e., mean
EPSC amplitude excluding failures) versus time (bin width, 33 s; n = 14). (F) Relative changes from control basal values of synaptic parameters before (basal) and after ND in control (n = 14),
AM251 (n = 6), BAPTA-loaded neuron (n = 11) and in hippocampal slices from transgenic mice lacking CB1Rs (n = 13). In E and F, only synapses that displayed e-SP in control were considered, except in BAPTA-loaded neurons and CB1R
-/-
, in which all recorded synapses were considered because previous identification of synapses in control was unfeasible (see Text). (G) Proportion of synapses that showed e-SP after ND in control (n = 51), in BAPTA-loaded neurons (n = 11), and
CB1R
-/-
(n = 13). **P < 0.01 and ***P < 0.001. Error bars indicate SEM.
Figure 2. Endocannabinoids depress transmitter release at hippocampal CA3-CA1 synapses.
(A) Schematic drawing depicting paired recordings from pyramidal neurons and the stimulating electrode. ND, neuronal depolarization. (B) Responses evoked by minimal stimulation (20

29 consecutive stimuli) (top traces) and averaged EPSCs (n = 50 stimuli; bottom traces) before (basal) and after ND in control and in 2 μM AM251. (C) EPSC amplitudes from a representative synapse.
Zero time corresponds to the beginning of ND. (D) Synaptic efficacy, Pr and synaptic potency versus time (n = 14). (E) Relative changes from control basal values of synaptic parameters before
(basal) and after ND in control (n = 14), AM251 (n = 9), BAPTA-loaded neuron (n = 11) and in hippocampal slices from transgenic mice lacking CB1Rs (n = 13). In D and E, only synapses that displayed DSE in control were considered, except in BAPTA-loaded neurons and CB1R
-/-
, in which all recorded synapses were considered because previous identification of synapses in control was unfeasible (see Text). (F) Proportion of synapses that showed DSE after ND in control (n = 51), in
BAPTA-loaded neurons (n = 11), and CB1R
-/-
(n = 13). **P < 0.01 and ***P < 0.001. Error bars indicate SEM.
Figure 3. Endocannabinoid-induced synaptic modulation is triggered by action potentials and depends on the neuronal activity level.
(A) Responses evoked by minimal stimulation (20 consecutive stimuli) (top traces) and averaged
EPSCs (n = 50 stimuli; bottom traces) before (basal) and after evoking trains of action potentials in current clamp conditions in two different synapses (left and right). (B) EPSC amplitudes from representative synapses. Zero time corresponds to the beginning of AP train. (C) Pr versus time in different synapses that displayed either e-SP (left; n = 14) or DSE (right; n = 12) in control. (D and
E) Relative changes from control basal values of synaptic parameters before (basal) and after APs in synapses that displayed either e-SP (n = 5) or DSE (n = 6), respectively, in control and after perfusion with AM251. Synapses displaying e-SP or DSE were identified in control. (F) Relative changes from control basal values of Pr versus number of APs elicited by stimulation of pyramidal neurons with depolarizing pulses of variable durations that evoked trains of APs (inset) in current

30 clamp conditions. Synapses were identified as showing DSE (open circles) or e-SP (filled circles).
*P < 0.05, **P < 0.01 and ***P < 0.001. Each value represents mean ± SEM of at least 4 synapses.
Figure 4. Intracellular Ca 2+ mobilization is necessary for the endocannabinoid-induced synaptic potentiation.
(A) Pseudocolor images representing fluorescence intensities of fluo 4-filled astrocytes before
(basal) and after ND in control and thapsigargin. Scale bar, 20
 m. (B) Ca
2+
levels from six astrocytes in a field of view and corresponding averaged traces (red) in control and thapsigargin.
Horizontal bars indicate ND. (C) Astrocyte Ca
2+
spike probability in control and thapsigargin (44 astrocytes from n = 7 slices). Zero time corresponds to the beginning of ND. (D) Mean EPSCs (n =
50) evoked by minimal stimulation before (basal) and after ND in control and thapsigargin. (E)
EPSC amplitudes from a representative synapse in control and thapsigargin. (F) Relative changes from control basal values of synaptic parameters before (basal) and after ND in synapses that displayed e-SP in control and thapsigargin (n = 7). *P < 0.05. Error bars indicate SEM (G)
Proportion of synapses that showed DSE or e-SP after ND in control and thapsigargin (n = 17). ***
P < 0.001
Figure 5. Endocannabinoid-induced synaptic potentiation requires astrocyte Ca 2+ elevations.
(A) Infrared DIC image showing paired recorded neurons, recorded astrocyte (bottom, left) and the stimulation electrode (bottom, right). (B) Schematic drawing depicting BAPTA dialysis into the astrocytic network from the recorded astrocyte. (C) (Left) Images representing fluorescence intensities of astrocytes before and after ND in BAPTA-loaded astrocytes. (Right) Astrocyte Ca 2+ spike probability in BAPTA-filled astrocytes (116 astrocytes from n = 15 slices). Zero time corresponds to the beginning of ND. (D) Synaptic efficacy, Pr and synaptic potency in slices with
BAPTA-filled astrocytes (n = 9). (E) (Left) Images representing fluorescence intensities of

31 astrocytes before and after ND in GDPβS-loaded astrocytes. (Right) Astrocyte Ca 2+ spike probability in GDPβS-filled astrocytes (146 astrocytes, n = 16 slices). (F) Synaptic efficacy, Pr and synaptic potency in slices with GDPβS-filled astrocytes (n = 10). (G) Relative changes from control basal values of synaptic parameters after ND in slices with BAPTA- (n = 9) and GDPβS-loaded astrocytes (n = 10). In D, F, and G, only synapses that did not display DSE were considered because previous identification of synapses in control was unfeasible (see Text). (H) Proportion of synapses that showed DSE or e-SP after ND in control (n = 51) and in slices with BAPTA- (n = 15) and
GDPβS-loaded astrocytes (n = 16). *** P < 0.001. Error bars indicate SEM. Scale bars, 20  m.
Figure 6. Endocannabinoid-evoked astrocyte-induced synaptic potentiation is mediated by mGluR type 1.
(A) Pseudocolor images representing astrocyte fluorescence intensities before (basal) and after ND in control and in the presence of the mGluR type 1 antagonist LY367385. Scale bar, 20 μm (B)
Astrocyte Ca
2+
spike probability versus time in control and LY367385 (n = 6 slices). Zero time corresponds to the beginning of the ND. (C) Astrocyte Ca
2+
oscillation frequency before (basal) and after ND in control and LY367385 (n = 6 slices). (D) Responses evoked by minimal stimulation (20 consecutive stimuli; top traces) and averaged EPSCs (n = 50 stimuli; bottom traces) before (basal) and after ND in control and LY367385. (E) EPSC amplitudes from a representative synapse in control and LY367385. (F) Pr versus time in control and after perfusion with LY367385 (n = 6). (G and H) Relative changes from control basal values of synaptic parameters before (basal) and after
ND in control and in the presence of LY367385 (n = 6) and MPEP (n = 7), respectively. *P < 0.05,
**P < 0.01. Error bars indicate SEM. (I) Schematic drawing representing the endocannabinoidmediated neuron-astrocyte signaling. Endocannabinoids (ECB) released by a pyramidal neuron activate astrocytic CB1Rs, elevate intracellular Ca
2+
, and stimulate glutamate (Glu) release that increase neurotransmitter release by activation of presynaptic mGluRs in a neighboring synapse.

32
Figure 7. The sign of the endocannabinoid-induced synaptic modulation depends on specific intercellular signaling pathways.
(A) Proportion of synapses that showed DSE or e-SP in control (n = 16) and PTX-treated slices (n =
21). (B) Schematic drawing representing the signaling pathways that lead to DSE or e-SP. (C)
Synaptic responses (20 consecutive stimuli; top traces) and averaged EPSCs (n = 50 stimuli; bottom traces) before (basal) and after ND in control and after perfusion with WIN. (D) EPSC amplitudes from a representative synapse in control and WIN. (E) Relative changes from control basal values of synaptic parameters before (basal) and after ND in control and WIN (n = 7). *P < 0.05, **P <
0.01and ***P < 0.001. Error bars indicate SEM.
Figure 8. Endocannabinoids induce homoneuronal synaptic depression and heteroneuronal synaptic potentiation.
(A) Pr after ND (relative to basal values) versus the respective distance between the somas of the stimulating and recording neurons (n = 56 synapses). (B) Schematic drawing depicting experimental arrangement. Homoneuronal synapse (green) corresponds to the recorded synapse onto the depolarized neuron. Heteroneuronal synapse (blue) corresponds to the recorded neighbour synapse onto the unstimulated neuron. (C) Synaptic responses (20 consecutive stimuli; top traces) and averaged EPSCs (n = 50 stimuli; bottom traces) before (basal) and after ND in heteroneuronal
(left) and homoneuronal synapses (right). (D) EPSC amplitudes simultaneously recorded from two representative heteroneuronal and homoneuronal synapses. Zero time corresponds to the beginning of the ND. (E) Proportion of heteroneuronal and homoneruonal synapses that showed DSE and e-
SP after ND (n = 33 paired synapses). ***P < 0.001. (F) Scheme representing the endocannabinoidmediated signaling processes. Endocannabinoids (ECBs) released by the pyramidal neuron acts directly on the respective presynaptic terminal leading to DSE in homoneuronal synapses. ECBs

33 acting on astrocytic CB1Rs elevate intracellular Ca 2+ , stimulate glutamate release and potentiate neurotransmitter release in heteroneuronal synapses.